Abstract

In his article “Reflexive Predictions” in a recent issue of this Journal [1], Professor Roger C. Buck refers to our theorem concerning correct public prediction of economic and, in general, social events [2]. Unfortunately he misunderstands the nature of our theorem. He begins by stipulating the infinite regress which originally led to the belief that correct public prediction in the social sciences might be impossible: the forecaster who attempts to take into account the agent's reaction to his public prediction, finds his adjusted public prediction again falsified by the agent's reaction to it, and so on ad infinitum [3]. Professor Buck then makes an admittedly tempting mistake. He argues that, provided the sequence of public predictions—each one adjusted for the agent's reaction to its predecessor—converges towards some value, correct public prediction is possible: at the end of a sequence of falsified predictions the forecaster, as it were, catches up with the agent.